Browse Prior Art Database

SYMBOLIC REGRESSION FOR BIG DATA DRIVEN PHYSICS MODELING

IP.com Disclosure Number: IPCOM000250347D
Publication Date: 2017-Jul-04
Document File: 4 page(s) / 673K

Publishing Venue

The IP.com Prior Art Database

Abstract

A symbolic regression analysis for big data driven physics model is disclosed. Symbolic regression searches the space of mathematical expressions to find the model that best fits a given dataset, both in terms of accuracy and simplicity. Input model comprises dataset and operator or function set, whereas output model is an equation that defines function from input features to an output value. However, genetic programming (GP) is used to solve the symbolic non-linear regression optimization problems. Symbolic regression requires relatively less theoretical knowledge to derive complex physics based models to best fit the observed lab or field data. Symbolic regression provides generic solution to derive models for different fields and assets.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

 SYMBOLIC REGRESSION FOR BIG DATA DRIVEN PHYSICS MODELING

BACKGROUND

 

The present disclosure relates generally to physics based models and more particularly to symbolic regression for big data driven physics modeling.

Physics based models are represented by complex mathematical formulas. Physics based models require theoretical field knowledge including different chemical and transport reactions, and empirical correlations with laboratory data, field data and mechanistic modeling. Further, physics based models are developed and evaluated manually, thereby making models difficult to generalize the process. Hence, varied results can be obtained for the same cases as different philosophies are used.

A large number of physics based models are developed for oil and gas industries. In 1973, de Waard and Milliams proposed a mechanistic model as follows:

Corrosion prediction is highly important for oil and gas industry. However, corrosion costs 2.2 trillion dollar per year to the industry, corresponding to 3.5 percent of global gross domestic product (GDP). Therefore, affecting the need for safeguarding the environment, personnel, and the cost associated with oil and gas production.

It would be desirable to have an efficient symbolic regression analysis for big data driven physics model.

BRIEF DESCRIPTION OF DRAWINGS

Figure 1 depicts an example of genetic programming (GP) used to solve the symbolic non-linear regression optimization problem.

Figure 2 depicts three types of physic models namely transport model, mechanistic model and operating model.

Figure 3 depicts a graphical representation between ground truth and symbolic regression.

Figure 4 depicts a graphical representation for symbolic regression analysis based on measurements suggesting corrosion.

DETAILED DESCRIPTION

A symbolic regression analysis for big data driven physics model is disclosed. Symbolic regression searches the space of mathematical expressions to find the model that best fits a given dataset, both in terms of accuracy and simplicity. As depicted in figure 1, input model comprises dataset and operator or function set, whereas output model is an equation that defines function from input features to an output value.  However, genetic programming (GP) is used to solve the symbolic non-linear regression optimization problems.

Figure 1

Symbolic regression requires relatively less theoretical knowledge to derive complex physics based models to best fit the observed lab...